There's a bug report that function library fails to compile with -fno-exceptions, with a patch attached: https://svn.boost.org/trac/boost/ticket/2900 The patch seems pretty safe -- any objections if I apply it? I primarily care because there is a similar bug report against program_options, which I can't fix until function is similarly modified. Of course, there's still the approach to declare that we don't care about -fno-exceptions at all. - Volodya